A method is investigated to estimate the full-field deformation spectra of a vibrating clamped plate using a set of PVDF strain sensors. A thin, fully clamped brass panel is instrumented with five piezo film sensors and excited using broadband noise from a loudspeaker. In an initial training stage, the panel vibration is measured simultaneously using the PVDF sensors and Digital Image Correlation. A linear mixing matrix is then learned that relates the sensor voltages to the modal coordinates of the first three structural modes. The number of time steps used for training should exceed ten times the number of PVDF sensors. The mixing matrix is then used to estimate full-field deformation spectra of another test with different excitation. The estimated resonant peaks show good agreement when compared to the true vibration peaks obtained from Digital Image Correlation and exhibit a higher signal-to-noise ratio than the DIC data.